Better distinction between IOException and TikaException

As discussed on the mailing list (http://markmail.org/message/qspwa2nqq5fksccs), many parser libraries throw IOExceptions even for errors that are not caused by problems reading bytes from the given document input stream. Tika should do a better job of catching such exceptions and converting them to TikaExceptions to better meet the Parser interface contract.

In Commons IO I just added a TaggedInputStream class (see IO-192) that is designed for better handling such cases.